crypto: tcrypt - reduce stack size

PATCH Superseded Frank Seidel 7 participants 22 messages Feb 25, 2009 – Mar 29, 2009 Crypto API

lore

Share & tools

lore

Frank Seidel

author

2009-02-25 13:48 UTC #0 lore raw

PATCH 1/1 crypto: tcrypt - reduce stack size

From: Frank Seidel <[email protected]>

Applying kernel janitors todos (printk calls need KERN_* constants on linebeginnings, reduce stack footprint where possible) to tcrypts test_hash_speed (where stacks memory footprint was very high (on i386 1184 bytes to 292 now).

Signed-off-by Frank Seidel <[email protected]>

---
 crypto/tcrypt.c |   29 ++++++++++++++++++++---------
 1 file changed, 20 insertions(+), 9 deletions(-)

crypto/tcrypt.c +20 -9

@@ -396,26 +396,34 @@ static void test_hash_speed(const char *
 	struct scatterlist sg[TVMEMSIZE];
 	struct crypto_hash *tfm;
 	struct hash_desc desc;
-	char output[1024];
+	char *output;
+	size_t output_size = 1024;
 	int i;
 	int ret;
 
-	printk("\ntesting speed of %s\n", algo);
+	printk(KERN_INFO "\ntesting speed of %s\n", algo);
+
+	output = kmalloc(output_size * sizeof(*output), GFP_KERNEL);
+	if (!output) {
+		printk(KERN_ERR "tcrypt: failed to allocate outputbuffer\n");
+		return;
+	}
 
 	tfm = crypto_alloc_hash(algo, 0, CRYPTO_ALG_ASYNC);
 
 	if (IS_ERR(tfm)) {
-		printk("failed to load transform for %s: %ld\n", algo,
+		printk(KERN_ERR "failed to load transform for %s: %ld\n", algo,
 		       PTR_ERR(tfm));
+		kfree(output);
 		return;
 	}
 
 	desc.tfm = tfm;
 	desc.flags = 0;
 
-	if (crypto_hash_digestsize(tfm) > sizeof(output)) {
-		printk("digestsize(%u) > outputbuffer(%zu)\n",
-		       crypto_hash_digestsize(tfm), sizeof(output));
+	if (crypto_hash_digestsize(tfm) > output_size) {
+		printk(KERN_ERR "digestsize(%u) > outputbuffer(%zu)\n",
+		       crypto_hash_digestsize(tfm), output_size);
 		goto out;
 	}

@@ -427,12 +435,14 @@ static void test_hash_speed(const char *
 
 	for (i = 0; speed[i].blen != 0; i++) {
 		if (speed[i].blen > TVMEMSIZE * PAGE_SIZE) {
-			printk("template (%u) too big for tvmem (%lu)\n",
+			printk(KERN_ERR
+			       "template (%u) too big for tvmem (%lu)\n",
 			       speed[i].blen, TVMEMSIZE * PAGE_SIZE);
 			goto out;
 		}
 
-		printk("test%3u (%5u byte blocks,%5u bytes per update,%4u updates): ",
+		printk(KERN_INFO "test%3u "
+		       "(%5u byte blocks,%5u bytes per update,%4u updates): ",
 		       i, speed[i].blen, speed[i].plen, speed[i].blen / speed[i].plen);
 
 		if (sec)

@@ -443,13 +453,14 @@ static void test_hash_speed(const char *
 					       speed[i].plen, output);
 
 		if (ret) {
-			printk("hashing failed ret=%d\n", ret);
+			printk(KERN_ERR "hashing failed ret=%d\n", ret);
 			break;
 		}
 	}
 
 out:
 	crypto_free_hash(tfm);
+	kfree(output);
 }
 
 static void test_available(void)

Thiago Galesi

2009-02-25 14:06 UTC #1 lore raw

Quoted 5 lines

int i;
int ret;
- printk("\ntesting speed of %s\n", algo);
+ printk(KERN_INFO "\ntesting speed of %s\n", algo);

Wouldn't it be better to take the first \n out as well??

Neil Horman

2009-02-25 14:06 UTC #2 lore raw

In reply to Frank Seidel

Quoted 11 lines · Frank Seidel

From: Frank Seidel <[email protected]>
Applying kernel janitors todos (printk calls need KERN_* constants on linebeginnings, reduce stack footprint where possible) to tcrypts test_hash_speed (where stacks memory footprint was very high (on i386 1184 bytes to 292 now).
Signed-off-by Frank Seidel <[email protected]>
---

crypto/tcrypt.c static void test_hash_speed(const char * · line 418

Quoted 6 lines

crypto/tcrypt.c +2 -1

 	if (IS_ERR(tfm)) {
-		printk("failed to load transform for %s: %ld\n", algo,
+		printk(KERN_ERR "failed to load transform for %s: %ld\n", algo,
 		       PTR_ERR(tfm));
+		kfree(output);
 		return;

Its just a stylistic nit, but I think I'd rather see this be a goto to the kfree at the end of the function. You can move the out label down to just the kfree, and rename the current out usage to out_free_tfm. I think if you're going to add the kfree at the bottom, you might as well make use of it for all cases.

If you change that, I'll ack this.

Thanks & Regards
Neil

Quoted 48 lines

crypto/tcrypt.c +9 -6

 	}
 
 	desc.tfm = tfm;
 	desc.flags = 0;
 
-	if (crypto_hash_digestsize(tfm) > sizeof(output)) {
-		printk("digestsize(%u) > outputbuffer(%zu)\n",
-		       crypto_hash_digestsize(tfm), sizeof(output));
+	if (crypto_hash_digestsize(tfm) > output_size) {
+		printk(KERN_ERR "digestsize(%u) > outputbuffer(%zu)\n",
+		       crypto_hash_digestsize(tfm), output_size);
 		goto out;
 	}

@@ -427,12 +435,14 @@ static void test_hash_speed(const char *
 
 	for (i = 0; speed[i].blen != 0; i++) {
 		if (speed[i].blen > TVMEMSIZE * PAGE_SIZE) {
-			printk("template (%u) too big for tvmem (%lu)\n",
+			printk(KERN_ERR
+			       "template (%u) too big for tvmem (%lu)\n",
 			       speed[i].blen, TVMEMSIZE * PAGE_SIZE);
 			goto out;
 		}
 
-		printk("test%3u (%5u byte blocks,%5u bytes per update,%4u updates): ",
+		printk(KERN_INFO "test%3u "
+		       "(%5u byte blocks,%5u bytes per update,%4u updates): ",
 		       i, speed[i].blen, speed[i].plen, speed[i].blen / speed[i].plen);
 
 		if (sec)

@@ -443,13 +453,14 @@ static void test_hash_speed(const char *
 					       speed[i].plen, output);
 
 		if (ret) {
-			printk("hashing failed ret=%d\n", ret);
+			printk(KERN_ERR "hashing failed ret=%d\n", ret);
 			break;
 		}
 	}
 
 out:
 	crypto_free_hash(tfm);
+	kfree(output);
 }
 
 static void test_available(void)

Herbert Xu

maintainer

2009-02-25 14:20 UTC #3 lore raw

In reply to Frank Seidel

Quoted 16 lines · Frank Seidel

From: Frank Seidel <[email protected]>
Applying kernel janitors todos (printk calls need KERN_* constants on linebeginnings, reduce stack footprint where possible) to tcrypts test_hash_speed (where stacks memory footprint was very high (on i386 1184 bytes to 292 now).
Signed-off-by Frank Seidel <[email protected]>
---
 crypto/tcrypt.c |   29 ++++++++++++++++++++---------
 1 file changed, 20 insertions(+), 9 deletions(-)

crypto/tcrypt.c static void test_hash_speed(const char * · line 399

Quoted 4 lines

crypto/tcrypt.c +0 -1

@@ -396,26 +396,34 @@ static void test_hash_speed(const char *
 	struct scatterlist sg[TVMEMSIZE];
 	struct crypto_hash *tfm;
 	struct hash_desc desc;
-	char output[1024];

How about just making it static?

Thanks,

Signature · 4 lines

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Frank Seidel

author

2009-02-25 14:24 UTC #4 lore raw

Quoted 3 lines · Herbert Xu

crypto/tcrypt.c static void test_hash_speed(const char * · line 399
Quoted 1 line
crypto/tcrypt.c +0 -1
-	char output[1024];
How about just making it static?

Well, it was static before. But there
are efforts to reduce current stack memory
footprints.

Frank

Herbert Xu

maintainer

2009-02-25 14:27 UTC #5 lore raw

Quoted 8 lines · Frank Seidel

Quoted 3 lines · Herbert Xu
crypto/tcrypt.c static void test_hash_speed(const char * · line 399
Quoted 1 line
crypto/tcrypt.c +0 -1
-	char output[1024];
How about just making it static?
Well, it was static before. But there
are efforts to reduce current stack memory
footprints.

Well if it's static it won't be using any stack memory.

Cheers,

Signature · 4 lines

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Geert Uytterhoeven

2009-02-25 14:31 UTC #6 lore raw

Quoted 11 lines · Herbert Xu

Quoted 8 lines · Frank Seidel
Quoted 3 lines · Herbert Xu
crypto/tcrypt.c static void test_hash_speed(const char * · line 399
Quoted 1 line
crypto/tcrypt.c +0 -1
-	char output[1024];
How about just making it static?
Well, it was static before. But there
are efforts to reduce current stack memory
footprints.
Well if it's static it won't be using any stack memory.

Probably people should be reminded tcrypt can only be a module, not built-in, so it won't waste precious unswappable kernel memory if it's static.

With kind regards,

Geert Uytterhoeven
Software Architect

Sony Techsoft Centre Europe
The Corporate Village · Da Vincilaan 7-D1 · B-1935 Zaventem · Belgium

Phone:    +32 (0)2 700 8453
Fax:      +32 (0)2 700 8622
E-mail:   [email protected]
Internet: http://www.sony-europe.com/

A division of Sony Europe (Belgium) N.V.
VAT BE 0413.825.160 · RPR Brussels
Fortis · BIC GEBABEBB · IBAN BE41293037680010
--
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Frank Seidel

author

2009-02-25 14:36 UTC #7 lore raw

Quoted 2 lines · Geert Uytterhoeven

Probably people should be reminded tcrypt can only be a module, not built-in, so it won't waste precious unswappable kernel memory if it's static.

Thats right, but it also doesn't hurt to simply use kmalloc, does it? :-) I just stumbled over tcrypt on the make checkstack output and as also the kerneljanitors todo advises to reduce this footprint where possible i just wanted to help out here.

Thanks,
Frank

Geert Uytterhoeven

2009-02-25 14:44 UTC #8 lore raw

Quoted 5 lines · Frank Seidel

Quoted 2 lines · Geert Uytterhoeven
Probably people should be reminded tcrypt can only be a module, not built-in, so it won't waste precious unswappable kernel memory if it's static.
Thats right, but it also doesn't hurt to simply use kmalloc, does it? :-)

Wel...

Using kmalloc() increases code size, makes the code more complex, and increases the risk of introducing a memory leak now or later.

Quoted 3 lines

I just stumbled over tcrypt on the make checkstack output and as also the kerneljanitors todo advises to reduce this footprint where possible i just wanted to help out here.

Reducing stack usage is fine. However, for a loadable test module without concurrency issues it's far easier to do that by just making the data static.

With kind regards,

Geert Uytterhoeven
Software Architect

Sony Techsoft Centre Europe
The Corporate Village · Da Vincilaan 7-D1 · B-1935 Zaventem · Belgium

Phone:    +32 (0)2 700 8453
Fax:      +32 (0)2 700 8622
E-mail:   [email protected]
Internet: http://www.sony-europe.com/

Frank Seidel

author

2009-02-25 14:54 UTC #9 lore raw

Quoted 5 lines · Geert Uytterhoeven

On Wed, 25 Feb 2009, Frank Seidel wrote:
Wel...
Using kmalloc() increases code size, makes the code more complex, and increases the risk of introducing a memory leak now or later.

Ok, admitted.

Quoted 6 lines

Quoted 3 lines
I just stumbled over tcrypt on the make checkstack output and as also the kerneljanitors todo advises to reduce this footprint where possible i just wanted to help out here.
Reducing stack usage is fine. However, for a loadable test module without concurrency issues it's far easier to do that by just making the data static.

Is PATCHv3 then ok for you?

Thanks,
Frank

Thiago Galesi

2009-02-25 14:29 UTC #10 lore raw

In reply to Frank Seidel

Quoted 8 lines · Frank Seidel

Quoted 3 lines · Herbert Xu
Quoted 1 line
- char output[1024];
How about just making it static?
Well, it was static before. But there
are efforts to reduce current stack memory
footprints.

If you write static char output[1024]; (even inside a function) it's not allocated on the stack.

Frank Seidel

author

2009-02-25 14:34 UTC #11 lore raw

Quoted 2 lines · Thiago Galesi

If you write static char output[1024]; (even inside a function) it's not allocated on the stack.

Oh, yes i misunderstood Herbert, sorry. But anyway isn't it preferred to kmalloc such arrays?
Greg, i thought it was you who told me so, is that right?

Thanks,
Frank

Greg KH

2009-02-25 19:47 UTC #12 lore raw

Quoted 7 lines · Frank Seidel

Quoted 2 lines · Thiago Galesi
If you write static char output[1024]; (even inside a function) it's not allocated on the stack.
Oh, yes i misunderstood Herbert, sorry. But anyway isn't it preferred to kmalloc such arrays?
Greg, i thought it was you who told me so, is that right?

Generally, yes, it is preferred. But if it can be static, that's fine too.

thanks,

greg k-h

Frank Seidel

author

2009-02-25 14:24 UTC #13 lore raw

PATCH v2 1/1 crypto: tcrypt - reduce stack size

In reply to Frank Seidel

From: Frank Seidel <[email protected]>

Signed-off-by Frank Seidel <[email protected]>

---
 crypto/tcrypt.c |   37 ++++++++++++++++++++++++-------------
 1 file changed, 24 insertions(+), 13 deletions(-)

crypto/tcrypt.c +24 ~~-13~~

@@ -396,27 +396,34 @@ static void test_hash_speed(const char *
 	struct scatterlist sg[TVMEMSIZE];
 	struct crypto_hash *tfm;
 	struct hash_desc desc;
-	char output[1024];
+	char *output;
+	size_t output_size = 1024;
 	int i;
 	int ret;
 
-	printk("\ntesting speed of %s\n", algo);
+	printk(KERN_INFO "\ntesting speed of %s\n", algo);
+
+	output = kmalloc(output_size * sizeof(*output), GFP_KERNEL);
+	if (!output) {
+		printk(KERN_ERR "tcrypt: failed to allocate outputbuffer\n");
+		return;
+	}
 
 	tfm = crypto_alloc_hash(algo, 0, CRYPTO_ALG_ASYNC);
 
 	if (IS_ERR(tfm)) {
-		printk("failed to load transform for %s: %ld\n", algo,
+		printk(KERN_ERR "failed to load transform for %s: %ld\n", algo,
 		       PTR_ERR(tfm));
-		return;
+		goto out;
 	}
 
 	desc.tfm = tfm;
 	desc.flags = 0;
 
-	if (crypto_hash_digestsize(tfm) > sizeof(output)) {
-		printk("digestsize(%u) > outputbuffer(%zu)\n",
-		       crypto_hash_digestsize(tfm), sizeof(output));
-		goto out;
+	if (crypto_hash_digestsize(tfm) > output_size) {
+		printk(KERN_ERR "digestsize(%u) > outputbuffer(%zu)\n",
+		       crypto_hash_digestsize(tfm), output_size);
+		goto out_free_tfm;
 	}
 
 	sg_init_table(sg, TVMEMSIZE);

@@ -427,12 +434,14 @@ static void test_hash_speed(const char *
 
 	for (i = 0; speed[i].blen != 0; i++) {
 		if (speed[i].blen > TVMEMSIZE * PAGE_SIZE) {
-			printk("template (%u) too big for tvmem (%lu)\n",
+			printk(KERN_ERR
+			       "template (%u) too big for tvmem (%lu)\n",
 			       speed[i].blen, TVMEMSIZE * PAGE_SIZE);
-			goto out;
+			goto out_free_tfm;
 		}
 
-		printk("test%3u (%5u byte blocks,%5u bytes per update,%4u updates): ",
+		printk(KERN_INFO "test%3u "
+		       "(%5u byte blocks,%5u bytes per update,%4u updates): ",
 		       i, speed[i].blen, speed[i].plen, speed[i].blen / speed[i].plen);
 
 		if (sec)

@@ -443,13 +452,15 @@ static void test_hash_speed(const char *
 					       speed[i].plen, output);
 
 		if (ret) {
-			printk("hashing failed ret=%d\n", ret);
+			printk(KERN_ERR "hashing failed ret=%d\n", ret);
 			break;
 		}
 	}
 
-out:
+out_free_tfm:
 	crypto_free_hash(tfm);
+out:
+	kfree(output);
 }
 
 static void test_available(void)

Frank Seidel

author

2009-02-25 14:53 UTC #14 lore raw

PATCH v3 1/1 crypto: tcrypt - reduce stack size

From: Frank Seidel <[email protected]>

Signed-off-by Frank Seidel <[email protected]>

---
 crypto/tcrypt.c |   16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

crypto/tcrypt.c +9 -7

@@ -396,16 +396,16 @@ static void test_hash_speed(const char *
 	struct scatterlist sg[TVMEMSIZE];
 	struct crypto_hash *tfm;
 	struct hash_desc desc;
-	char output[1024];
+	static char output[1024];
 	int i;
 	int ret;
 
-	printk("\ntesting speed of %s\n", algo);
+	printk(KERN_INFO "\ntesting speed of %s\n", algo);
 
 	tfm = crypto_alloc_hash(algo, 0, CRYPTO_ALG_ASYNC);
 
 	if (IS_ERR(tfm)) {
-		printk("failed to load transform for %s: %ld\n", algo,
+		printk(KERN_ERR "failed to load transform for %s: %ld\n", algo,
 		       PTR_ERR(tfm));
 		return;
 	}

@@ -414,7 +414,7 @@ static void test_hash_speed(const char *
 	desc.flags = 0;
 
 	if (crypto_hash_digestsize(tfm) > sizeof(output)) {
-		printk("digestsize(%u) > outputbuffer(%zu)\n",
+		printk(KERN_ERR "digestsize(%u) > outputbuffer(%zu)\n",
 		       crypto_hash_digestsize(tfm), sizeof(output));
 		goto out;
 	}

@@ -427,12 +427,14 @@ static void test_hash_speed(const char *
 
 	for (i = 0; speed[i].blen != 0; i++) {
 		if (speed[i].blen > TVMEMSIZE * PAGE_SIZE) {
-			printk("template (%u) too big for tvmem (%lu)\n",
+			printk(KERN_ERR
+			       "template (%u) too big for tvmem (%lu)\n",
 			       speed[i].blen, TVMEMSIZE * PAGE_SIZE);
 			goto out;
 		}
 
-		printk("test%3u (%5u byte blocks,%5u bytes per update,%4u updates): ",
+		printk(KERN_INFO "test%3u "
+		       "(%5u byte blocks,%5u bytes per update,%4u updates): ",
 		       i, speed[i].blen, speed[i].plen, speed[i].blen / speed[i].plen);
 
 		if (sec)

@@ -443,7 +445,7 @@ static void test_hash_speed(const char *
 					       speed[i].plen, output);
 
 		if (ret) {
-			printk("hashing failed ret=%d\n", ret);
+			printk(KERN_ERR "hashing failed ret=%d\n", ret);
 			break;
 		}
 	}

Neil Horman

Acked-by on v3 1/1

2009-02-25 15:32 UTC #15 lore raw

Quoted 9 lines · Frank Seidel

From: Frank Seidel <[email protected]>
Applying kernel janitors todos (printk calls need KERN_* constants on linebeginnings, reduce stack footprint where possible) to tcrypts test_hash_speed (where stacks memory footprint was very high (on i386 1184 bytes to 160 now).
Signed-off-by Frank Seidel <[email protected]>

Looks good, thanks Frank

Acked-by Neil Horman <[email protected]>

Herbert Xu

maintainer

2009-03-29 07:20 UTC #16 lore raw

Quoted 9 lines · Neil Horman

Quoted 8 lines · Frank Seidel
From: Frank Seidel <[email protected]>
Applying kernel janitors todos (printk calls need KERN_* constants on linebeginnings, reduce stack footprint where possible) to tcrypts test_hash_speed (where stacks memory footprint was very high (on i386 1184 bytes to 160 now).

Quoted 3 lines

Quoted 1 line
Signed-off-by Frank Seidel <[email protected]>
Looks good, thanks Frank
Acked-by Neil Horman <[email protected]>

Patch applied to cryptodev. Thanks!

Signature · 4 lines

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Geert Uytterhoeven

2009-02-25 15:54 UTC #17 lore raw

In reply to Frank Seidel

crypto/tcrypt.c static void test_hash_speed(const char * · line 403

Quoted 2 lines · Frank Seidel

crypto/tcrypt.c +1 -1

-	printk("\ntesting speed of %s\n", algo);
+	printk(KERN_INFO "\ntesting speed of %s\n", algo);

                          ^^
			  woops

BTW, why are you using printk(KERN_INFO "..."), and not pr_info("...")?

With kind regards,

Geert Uytterhoeven
Software Architect

Sony Techsoft Centre Europe
The Corporate Village · Da Vincilaan 7-D1 · B-1935 Zaventem · Belgium

Phone:    +32 (0)2 700 8453
Fax:      +32 (0)2 700 8622
E-mail:   [email protected]
Internet: http://www.sony-europe.com/

Frank Seidel

2009-02-25 15:56 UTC #18 lore raw

Quoted 1 line · Geert Uytterhoeven

BTW, why are you using printk(KERN_INFO "..."), and not pr_info("...")?

Without special reason. Is pr_info preferred?

Thanks,
Frank

Geert Uytterhoeven

2009-02-25 15:59 UTC #19 lore raw

Quoted 4 lines · Frank Seidel

Quoted 1 line · Geert Uytterhoeven
BTW, why are you using printk(KERN_INFO "..."), and not pr_info("...")?
Without special reason. Is pr_info preferred?

It's shorter.

Especially pr_err' fits in the space of printk', so you don't have to split lines that become to long due to adding the KERN_*.

With kind regards,

Geert Uytterhoeven
Software Architect

Sony Techsoft Centre Europe
The Corporate Village · Da Vincilaan 7-D1 · B-1935 Zaventem · Belgium

Phone:    +32 (0)2 700 8453
Fax:      +32 (0)2 700 8622
E-mail:   [email protected]
Internet: http://www.sony-europe.com/

Frank Seidel

author

2009-02-25 16:02 UTC #20 lore raw

Quoted 4 lines · Geert Uytterhoeven

It's shorter.
Especially pr_err' fits in the space of printk', so you don't have to split lines that become to long due to adding the KERN_*.

Ah, i see. Thanks for the explanation :-)

Frank

Geert Uytterhoeven

2009-02-25 16:09 UTC #21 lore raw

In reply to Geert Uytterhoeven

Quoted 10 lines · Geert Uytterhoeven

Quoted 4 lines · Frank Seidel
Quoted 1 line · Geert Uytterhoeven
BTW, why are you using printk(KERN_INFO "..."), and not pr_info("...")?
Without special reason. Is pr_info preferred?
It's shorter.
Especially pr_err' fits in the space of printk', so you don't have to split lines that become to long due to adding the KERN_*.

I sent it out too soon...

Other advantages:
- The pr_*() calls use pr_fmt(), so you can set default prefixes (like __FILE__, __function, __LINE) at one place at the top of your driver source file.
- pr_debug() is automatically left out if DEBUG is not defined (unless CONFIG_DYNAMIC_PRINTK_DEBUG is set).

With kind regards,

Geert Uytterhoeven
Software Architect

Sony Techsoft Centre Europe
The Corporate Village · Da Vincilaan 7-D1 · B-1935 Zaventem · Belgium

Phone:    +32 (0)2 700 8453
Fax:      +32 (0)2 700 8622
E-mail:   [email protected]
Internet: http://www.sony-europe.com/